Leaderboard
On-device LLM performance rankings powered by Glicko-2
iPhone 16 Pro Max
iOSRank
#9
Rating
1,966
±13 RD
Win Rate
95.0%
Conservative Rating
1,940
TG Rating
1,949
PP Rating
1,966
Matches
1,568
Record
1489W – 79L
Models Tested
| Model | TG Median (tok/s) | PP Median (tok/s) | TG Best | PP Best | Runs |
|---|---|---|---|---|---|
| DeepSeek-R1-Distill-Qwen-7B-Q4_K_M | 1385.49 | 78705.26 | 2767.97 | 157404.57 | 2 |
| tinygemma3-Q8_0 | 324.43 | 25507.49 | 324.43 | 25507.49 | 1 |
| SmolLM2-135M-Instruct-Q8_0 | 177.21 | 5002.24 | 179.21 | 5020.62 | 2 |
| Qwen3-0.6B-Q8_0 | 61.33 | 1240.41 | 61.33 | 1240.41 | 1 |
| DeepSeek-R1-Distill-Qwen-7B-Q4_K_L | 60.77 | 587.80 | 60.77 | 587.80 | 1 |
| Vintern-1B-v3_5-Q8_0 | 59.27 | 1473.72 | 59.27 | 1473.72 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-IQ3_XS | 47.56 | 514.04 | 47.56 | 514.04 | 1 |
| gemma-3-1b-it.Q3_K_M | 45.66 | 79.93 | 45.66 | 79.93 | 1 |
| gemma-3-1b-it-Q8_0 | 40.61 | 954.90 | 40.61 | 954.90 | 1 |
| gemma-3-1b-it.Q8_0 | 40.48 | 907.78 | 40.63 | 964.20 | 2 |
| llama-3.2-1b-instruct-q8_0 | 32.67 | 72.79 | 39.03 | 797.09 | 7 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M | 28.57 | 45.97 | 28.57 | 45.97 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q8_0 | 28.22 | 526.84 | 28.25 | 529.16 | 2 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M | 27.98 | 44.54 | 27.98 | 44.54 | 1 |
| qwen2.5-1.5b-instruct-q8_0 | 25.18 | 52.41 | 30.58 | 586.27 | 3 |
| Qwen2.5-Coder-3B-Instruct-abliterated-Q4_K_M | 24.34 | 261.23 | 24.34 | 261.23 | 1 |
| SmolLM2-1.7B-Instruct-Q8_0 | 24.18 | 427.16 | 24.39 | 430.03 | 2 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q8_0 | 23.37 | 155.27 | 26.02 | 287.07 | 3 |
| Gemmasutra-Mini-2B-v1-Q6_K | 20.97 | 287.40 | 22.24 | 346.63 | 7 |
| Qwen2.5-Coder-3B-Instruct-abliterated-Q5_K_M | 20.68 | 230.28 | 20.68 | 230.28 | 1 |
| deepscaler-1.5b-preview-q8_0 | 20.51 | 46.80 | 20.51 | 46.80 | 1 |
| Qwen3.5-2B-IQ4_NL | 20.36 | 340.57 | 20.36 | 340.57 | 1 |
| gemma-3-4b-it-IQ4_NL | 19.28 | 202.07 | 19.28 | 202.07 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q6_K_L | 19.27 | 252.92 | 33.94 | 456.20 | 4 |
| sonnet-llama-3.2-3b.Q4_K_M | 19.21 | 172.29 | 19.21 | 172.29 | 1 |
| gemma-3-4B-it-QAT-Q4_0 | 19.06 | 213.09 | 19.75 | 215.65 | 2 |
| bootes-qwen3_coder-reasoning-q4_k_m | 19.04 | 195.53 | 19.04 | 195.53 | 1 |
| Qwen3-4B-Instruct-2507-Q4_K_M | 18.93 | 197.58 | 18.93 | 197.58 | 1 |
| Hermes-3-Llama-3.2-3B.Q5_K_M | 18.80 | 206.48 | 18.80 | 206.48 | 1 |
| amoral-gemma3-4B-v2.IQ4_XS | 18.62 | 195.43 | 18.62 | 195.43 | 1 |
| LFM2-1.2B-F16 | 18.52 | 35.81 | 18.52 | 35.81 | 1 |
| gemma-3n-E2B-it-IQ4_XS | 18.50 | 29.66 | 18.50 | 29.66 | 1 |
| Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking_Q4_k_m | 18.48 | 198.21 | 18.48 | 198.21 | 1 |
| gemma-3-4b-it-Q4_K_M | 18.10 | 204.06 | 18.10 | 204.06 | 1 |
| gemma-3-4b-it-q4_0_s | 17.66 | 187.37 | 19.79 | 215.40 | 5 |
| EXAONE-3.5-2.4B-Instruct-Q8_0 | 17.31 | 263.20 | 17.31 | 263.20 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-Q8_0 | 16.54 | 51.46 | 16.54 | 51.46 | 1 |
| Qwen3-1.7B.Q6_K | 16.48 | 30.06 | 16.48 | 30.06 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-f16 | 15.75 | 520.05 | 15.75 | 520.05 | 1 |
| DeepSeek-R1-Distill-Qwen-1.5B-IQ2_M | 15.58 | 22.25 | 17.04 | 23.32 | 2 |
| Bootes-Qwen3_Coder-Reasoning.Q5_K_M | 15.20 | 174.98 | 15.20 | 174.98 | 1 |
| google_gemma-3-4b-it-Q6_K | 14.86 | 199.10 | 14.86 | 199.10 | 1 |
| llama-3.2-3b-instruct-q8_0 | 14.28 | 238.37 | 14.28 | 238.37 | 1 |
| gemma-3-4b-it.Q4_K_M | 14.27 | 110.03 | 18.02 | 201.75 | 2 |
| Phi-4-mini-reasoning-Q6_K | 14.23 | 183.22 | 14.23 | 183.22 | 1 |
| amoral-gemma3-4B-v2.Q4_K_M | 13.93 | 158.55 | 13.93 | 158.55 | 1 |
| gemma-3-4b-it-Q4_K_M | 13.76 | 104.50 | 17.88 | 191.64 | 2 |
| gemma-3n-E2B-it-Q4_1 | 13.53 | 22.66 | 13.53 | 22.66 | 1 |
| Llama-3.2-3B-Instruct-uncensored-Q8_0 | 13.46 | 218.65 | 14.08 | 232.68 | 2 |
| DeepSeek-R1-0528-Qwen3-8B-Q2_K | 13.26 | 94.94 | 13.26 | 94.94 | 1 |
1–50 of 101 rows
1 / 3
Head-to-Head Record
1–50 of 353 rows
1 / 8
Performance by App Version
ImprovedRegressed